Optimal quantisation of the discrete cosine transform for image compression

We propose a method for constructing optimal quantisation tables for use with discrete cosine transform (DCT) based image compressors. The method produces optimal rate distortion performance when linear quantisation and a fixed block size DCT are used. The quantisation table for the image is transmitted as part of the compressed file. We show that the rate distortion performance is superior to JPEG and close to that of wavelet methods. Unlike JPEG our method is progressive and produces the exact compression ratio required